SSE-To-Metadata Filter (proto)
This extension has the qualified name envoy.filters.http.sse_to_metadata
Note
This extension is functional but has not had substantial production burn time, use only with this caveat.
This extension is intended to be robust against untrusted downstream traffic. It assumes that the upstream is trusted.
Tip
This extension extends and can be used with the following extension category:
This extension must be configured with one of the following type URLs:
Warning
This API feature is currently work-in-progress. API features marked as work-in-progress are not considered stable, are not covered by the threat model, are not supported by the security team, and are subject to breaking changes. Do not use this feature without understanding each of the previous points.
The SSE-To-Metadata filter extracts values from Server-Sent Events (SSE) HTTP response bodies and writes them to dynamic metadata. This is useful for LLM token usage tracking, logging, and other observability use cases.
The filter specifically handles SSE format (text/event-stream) and uses pluggable content parsers to extract values from the SSE data fields. The content parser is a typed extension that can be configured to handle different content types (JSON, plaintext, XML, etc.).
The filter only processes responses with Content-Type “text/event-stream” (the standard SSE content type). Content-Type parameters such as charset are ignored.
See SSE-To-Metadata configuration overview for more details.
extensions.filters.http.sse_to_metadata.v3.SseToMetadata
[extensions.filters.http.sse_to_metadata.v3.SseToMetadata proto]
{
"response_rules": {...}
}
- response_rules
(extensions.filters.http.sse_to_metadata.v3.SseToMetadata.ProcessingRules, REQUIRED) Rules for processing SSE response streams.
extensions.filters.http.sse_to_metadata.v3.SseToMetadata.ProcessingRules
[extensions.filters.http.sse_to_metadata.v3.SseToMetadata.ProcessingRules proto]
Rules for processing SSE streams and extracting metadata.
The filter parses the SSE protocol (events delimited by blank lines), then delegates to a content parser to parse event content and extract metadata. The content parser determines which values to extract and how to write them to metadata.
{
"content_parser": {...},
"max_event_size": {...}
}
- content_parser
(config.core.v3.TypedExtensionConfig, REQUIRED) Content parser configuration for parsing event content and extracting metadata.
The content parser specifies: - How to parse the event data (e.g., JSON, XML, plaintext) - Which values to extract from the parsed content (e.g., JSON paths like usage.total_tokens) - How to map extracted values to metadata (namespace, key, type conversions) - When to write metadata (on_present, on_missing, on_error actions)
- max_event_size
(UInt32Value) Maximum size in bytes for a single SSE event before it’s considered invalid and discarded. This protects against unbounded memory growth from malicious or malformed streams that never send event delimiters (blank lines).
Default is 8192 bytes (8KB), which is sufficient for most legitimate events. Set to 0 to disable the limit (not recommended for production). Maximum allowed value is 10485760 bytes (10MB).